Someone please explain this upsampling BS to me (rant)

[Canman]

OK, I don't want this to start an argument here, but let's talk about the upsampling BS that's been going on.

The hot thing these days is to have "upsampling" to 192khz. I understand the differentation from oversampling that the folks around here like to make and I don't see one as being more technologically advanced than the other. Oversampling, or synchronous upsampling makes more sense to me due to the simpler mathematical algorithms involved.

But the real question I have is this: What's the big deal here? CD players have been using oversampling for years to make for a simpler low pass analog filter. A more recent advance is calculating new samples at 24 bits, but not even that is very new.

The Wadia 302 does synchonous upsampling of 16 times to 705.6kHz at 24 bits and the 27ix DAC does synchronous upsampling of 64 times at 2.82 mHz.

So why would asynchronous upsampling to 192kHz be a good thing, or even something you would want to advertise? I understand that most machines don't have sufficient processing power to decode anything beyond 24 bit, 192kHz (The Wadias run multiple DACs per channel to achieve suffucient processing power), so why would 24/192 become a marketing buzzword when to me there doesn't seem to be anything special about it?

End rant.

 

[Rameish]

I wont even try to expain the techinical issues for 2 reasons: 1) I'm not competent enough and 2) it doesnt solve the real question of what sounds better in your opinion (not mine). The best way to arrive at you preferred sound is to set aside some time and make an appointment with a reputable dealer in your area that has several brands of cdps/sacdp/dvd-a players and sit down with your favourite music and let your mind relax. Then when you have listened to the 2 or at best 3 players pick 2 and from there prick one. Reviews (done either by professionals or consumers) are just starting points - a shortlist if you like - but dont overlook your dealer's advice as well if he has a lesser known or lesser reviewd model it may just be your cuppa.

For more information of upsampling, oversampling and Non OverSampling visit www.diyaudio.com (digital section). Over there Non Over-Sampling (NOS for short over there) is all the rave. Looks like using a 16 bit chip like the TDA1541 (either A or Si or S2 version) and not oversampling it is the way to go to get a "natural" sound for redbook playback.

I've heard Upsampled CD players that sound bad and some that sond really good likewise for Oversampling. As always, looking at specs is misleading and isnt worth the paper it's printed on - as a guide.

But since many here come from a IT background I guess you can look at it - but dont forget to then toss it in the round file LOL.

 

[eyeteeth]

Upsampling, Upconversion and Oversampling:A Marketing Synonym Game at the Simaudio site is a fairly clear, low BS article. Although I'm sure they'd like to sell some of their products.

 

[JaZZ]

I think there's a sonic difference between asynchronous (upsampling) and synchronous (oversampling) upconversion -- and of course non-upsampling filtering. I have a DVD 963SA with three possible modes: non-upsampling, 96- and 192-kHz upsampling (both with 24 bit). Moreover my Bel Canto DAC2 has a fixed upsampling rate of 192 kHz. I would say there's a clear sonic relationship between the DAC2 and the Philips at 192 kHz, whereas 96 kHz is something between non-upsampling and 192 kHz. What upsampling does is increase the perception of transient accuracy and spatial depth, at the price of a somewhat edgier sound. After preferring 192 kHz in the beginning 96 kHz was my favorite for quite some time with the Philips.

I've always left open that the sonic benefits from upsampling could in fact be kind of euphonic distortion artifacts from the estimation process -- it's true, synchronous upconversion theoretically makes more sense --, but I liked them. And threre were some highly praised benchmark products such as Meitner and dCs which swear by upsampling, so how can it be wrong!

Well, currently I have a very highly resolving headphone amp for audition, and with this one I'm halfways convinced that upsampling makes the sound unnecessarily edgy and harsh -- it shows all its well-known merits through a not-so-high-end chain, but fails to please as soon as you have a chain that shows the last bit of signal information. Maybe that's a conclusion which again could be prooved wrong in the future, but that's how I hear it now. The DVD 963SA has never sounded sweeter than it does now, without upsampling.

 

[ITZBITZ]

Photoshop can resample an image to any size, but the resulting quality is determined by the quality of the source image. This same resampling technology is how audio is resampled.

Upsampling and oversampling are the same thing. Oversampling is done on integer multiples of the original sampling frequency to simply math back when chips were slow. Today, chips can crunch floating point like crazy. So an integer math solution is no longer required.

The reason chips now use non-integer sampling is simple. The pro-audio world uses 48Khz (and now 96Khz and 192Khz) and DVD players were made to use 96Khz audio. So much for integer resampling when playing a CD. Chances are the DAC can only take a single input at the hardware level based on the crystals provided. Many have two crystals, some only have one. Those with a single crystal have to resample the audio to convert it to analog.

Take two adjacent audio samples, one is 20000 and the next one is 21000. With a simple 2x upsample, you'll get a new 20500 value in between the previous two values. That's not the real problem. The problem is when you have two values, 20000 and 20000. The upsample results in an additional 20000 value being added. Except that were it recorded in 88.2Khz, it might have actually been 20000, 21000, 20000. A peak may have been lost due to sampling rate. There are plenty of interesting algorithms one can use to try and guess that it might not be a flat transition, but they are all approximations as there is NO way to retrieve the missing sample.

A non-upsampling (integer or otherwise) DAC has to use something to roll off the sampling noise above 20K. It's also highly likely the audio for your CD was recording in 48, 96 or 192Khz first and the resampled to 44.1 Khz. An oversampling DAC can use a high sampling frequency and a simple filter to prevent hash from entering the audible range. Thing is you can't hear 50 Khz, but your electronics can and it can introduce problems in your output/amplification stage if those high-frequency signals get into circuits designed to handle the audio band.

Okay, I'm rambling like crazy but understanding how things working inside the case can really help one realize how a process can affect the components that bring the sound to your ears.

 

[jefemeister]

Quote:

Originally Posted by Canman But the real question I have is this: What's the big deal here? CD players have been using oversampling for years to make for a simpler low pass analog filter.

It's simple. For the longest time, digital audio was strictly 44.1kHz (maybe 48kHz in some cases). Resampling was done then for the same reasons it's done today. However, today there are many different sample rates, and somehow the public has been tricked into thinking that there is something magical about the exact numbers of 96kHz or 192kHz since these are the "hi-rez" rates.

To further compound the problem, people have made up their minds about sample rates based on 1 or 2 products.
For example, a particular DAC may sound better fed with one particular rate than another. The decision is then made that "rate A sounds better than rate B." This may be valid for that particular DAC but it will not hold true for other DACs. But the assocation is already made. The fact that certain companies then release "sample rate converter boxes" and "resampling transports" and the like throws yet another wrench into the works.

Bottom line: It's all about IMPLEMENTATION.

 

[Canman]

Quote:

Originally Posted by jefemeister It's simple. For the longest time, digital audio was strictly 44.1kHz (maybe 48kHz in some cases). Resampling was done then for the same reasons it's done today. However, today there are many different sample rates, and somehow the public has been tricked into thinking that there is something magical about the exact numbers of 96kHz or 192kHz since these are the "hi-rez" rates. Bottom line: It's all about IMPLEMENTATION.

That is exactly my point. Upsampling to 96/192kHz makes it sound like it is Hi-Rez, when in fact the technology to do this is nothing new. If anything, that tells me that the DSP / digital filter method is off the shelf.

But it doesn't really matter. I have heard great CD playback using each of synchronous upsampling, asynchronous upsampling, or non-oversampling. As jefemeister said, its all about how you implement the chipset or circuit.

Sorry this thread was more of a rant than a search for new information. Hopefully someone got something out of this (Upsampling is not a new, magical method for CD playback)

 

[jefemeister]

Quote:

Originally Posted by Canman That is exactly my point. Upsampling to 96/192kHz makes it sound like it is Hi-Rez, when in fact the technology to do this is nothing new. If anything, that tells me that the DSP / digital filter method is off the shelf.

Well, to be a little fair, most DAC chips can't accept more than 24/192. So they're really doing the maximum amount of upsampling. The PCM1704 is special in that it can handle up to 24/768.

As a side note, the PCM1704 has a matching chip called the DF1704 that is a canned 8x upsampling filter. Some DAC chips integrate these types of filters directly into the DAC so that the DAC is *always* upsampling even if you don't think it is.

 

[AndreYew]

Sampling rate ratios that are not integral to each other (like 44.1 and 96) doesn't mean you can't do integral sample-rate conversion. For example, a 320x upsampling of 44.1 kHz followed by a 147x downsampling will get you to 96 kHz. Of course you have to do more computation, but it's still integral.

JaZZ's point of synchronous vs. asynchronous SRC is interesting as well from many perspectives. ASRC has to be able to accept any sample rate within its input spec at every sample received, which means its design is inherently compromised because it has to be a bit more flexible, whereas SSRC can be optimized for the particular input and output sample rates. ASRC sees jitter on its input as a sample rate change, and will convert jitter to different kinds of noise or artifacts, depending on its internal design. Most upsampling boxes use an ASRC chip. Done well, like Benchmark's DAC-1, it can be very effective (their anti-jitter circuit uses an ASRC), but I don't trust most audio companies to do digital design well, and I have to wonder how many upsampling boxes render an inaccurate conversion due to sensitivity to its input signal. I also have to wonder how much some of these companies that have ASRC know what they're doing, or whether they're just tossing it in as a flavor of the month. As Jeff says, it's all about the implementation.

Finally, Dan Lavry's written a long, but very informative, essay on why 192 kHz is a technical mistake:

http://www.lavryengineering.com/docu...ing_Theory.pdf

He produces very fine DACs and ADCs for the pro audio market.

--Andre

 

[fewtch]

Quote:

Originally Posted by AndreYew Finally, Dan Lavry's written a long, but very informative, essay on why 192 kHz is a technical mistake: http://www.lavryengineering.com/docu...ing_Theory.pdf He produces very fine DACs and ADCs for the pro audio market.

He probably doesn't understand the audiophile maxim that if enough is good, too much is better, and way too much even better still...

 

[eyeteeth]

Quote:

Originally Posted by AndreYew Finally, Dan Lavry's written a long, but very informative, essay on why 192 kHz is a technical mistake:

That is why Simaudio disregards the 96kHz and 196kHz rates.

SIMAUDIO CDPs
"In numerical terms this represents 24-bit/352.8kHz digital signal resolution, which is significantly higher than that of either a 24-bit/96kHz or 24-bit/192kHz upsampler. What would be the point of using a lower resolution technique? Furthermore, the 352.8kHz digital signal does not suffer from any error and mathematical truncation resulting from the use of a non-integer multiple sample rate conversion frequencies like 96kHz and 192kHz."

 

[AndreYew]

I think Dan believes 192 kHz isn't useful because the parts that can do 192 kHz conversion (specifically ADCs) compromise accuracy in order to run that fast. The paper has more details.

--Andre

 

[taylor]

My understanding is that it 'smooths' out the signal. I think this picture I made can explain it better than I can.

 

[Dusty Chalk]

I am basically going to stay out of this one, but I will ask the question I always ask when someone steps on the soapbox to say that they're the same thing -- isn't the advantage of increased bit depth worth it?

Ultimately, I do agree with jefemeister, in that implementation is everything, and you shouldn't make it a criterion in what you buy -- ultimately, one should listen with their ears, and decide on which source component is the most satisfactory for them.

I still want that Tri-Vista CD player.

 

[Zoide]

Wouldn't upsampling or oversampling simply mean that you are using algorithms to generate "extra" samples so that your sound is more similar to analog? As in, you'd get smooth waves with vinyl instead of jagged rectangles with digital recording, but with higher sampling rates the jaggedness is reduced and you get closer to the original, smooth waves?